Parent Directory
| 
Revision Log
Revision 1.33.2.23 - (view) (download)
| 1 : | edgomez | 1.16 | /***************************************************************************** |
| 2 : | edgomez | 1.17 | * |
| 3 : | * XVID MPEG-4 VIDEO CODEC | ||
| 4 : | * - Native API implementation - | ||
| 5 : | * | ||
| 6 : | * This program is an implementation of a part of one or more MPEG-4 | ||
| 7 : | * Video tools as specified in ISO/IEC 14496-2 standard. Those intending | ||
| 8 : | * to use this software module in hardware or software products are | ||
| 9 : | * advised that its use may infringe existing patents or copyrights, and | ||
| 10 : | * any such use would be at such party's own risk. The original | ||
| 11 : | * developer of this software module and his/her company, and subsequent | ||
| 12 : | * editors and their companies, will have no liability for use of this | ||
| 13 : | * software or modifications or derivatives thereof. | ||
| 14 : | * | ||
| 15 : | * This program is free software ; you can redistribute it and/or modify | ||
| 16 : | * it under the terms of the GNU General Public License as published by | ||
| 17 : | * the Free Software Foundation ; either version 2 of the License, or | ||
| 18 : | * (at your option) any later version. | ||
| 19 : | * | ||
| 20 : | * This program is distributed in the hope that it will be useful, | ||
| 21 : | * but WITHOUT ANY WARRANTY ; without even the implied warranty of | ||
| 22 : | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
| 23 : | * GNU General Public License for more details. | ||
| 24 : | * | ||
| 25 : | * You should have received a copy of the GNU General Public License | ||
| 26 : | * along with this program ; if not, write to the Free Software | ||
| 27 : | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
| 28 : | * | ||
| 29 : | ****************************************************************************/ | ||
| 30 : | chenm001 | 1.29 | |
| 31 : | edgomez | 1.16 | /***************************************************************************** |
| 32 : | edgomez | 1.17 | * |
| 33 : | * History | ||
| 34 : | * | ||
| 35 : | suxen_drol | 1.21 | * - 23.06.2002 added XVID_CPU_CHKONLY |
| 36 : | edgomez | 1.17 | * - 17.03.2002 Added interpolate8x8_halfpel_hv_xmm |
| 37 : | * - 22.12.2001 API change: added xvid_init() - Isibaar | ||
| 38 : | * - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> | ||
| 39 : | * | ||
| 40 : | syskin | 1.33.2.23 | * $Id: xvid.c,v 1.33.2.22 2003/01/04 06:14:32 suxen_drol Exp $ |
| 41 : | edgomez | 1.17 | * |
| 42 : | ****************************************************************************/ | ||
| 43 : | Isibaar | 1.1 | |
| 44 : | suxen_drol | 1.33.2.21 | #include <stdio.h> |
| 45 : | suxen_drol | 1.33.2.20 | #include <stdlib.h> |
| 46 : | suxen_drol | 1.33.2.21 | #include <string.h> |
| 47 : | suxen_drol | 1.33.2.20 | #include <time.h> |
| 48 : | |||
| 49 : | Isibaar | 1.1 | #include "xvid.h" |
| 50 : | #include "decoder.h" | ||
| 51 : | #include "encoder.h" | ||
| 52 : | #include "bitstream/cbp.h" | ||
| 53 : | #include "dct/idct.h" | ||
| 54 : | #include "dct/fdct.h" | ||
| 55 : | #include "image/colorspace.h" | ||
| 56 : | #include "image/interpolate8x8.h" | ||
| 57 : | suxen_drol | 1.33.2.15 | #include "image/reduced.h" |
| 58 : | Isibaar | 1.1 | #include "utils/mem_transfer.h" |
| 59 : | h | 1.33.2.5 | #include "utils/mbfunctions.h" |
| 60 : | Isibaar | 1.1 | #include "quant/quant_h263.h" |
| 61 : | #include "quant/quant_mpeg4.h" | ||
| 62 : | ia64p | 1.30 | #include "motion/motion.h" |
| 63 : | Isibaar | 1.1 | #include "motion/sad.h" |
| 64 : | #include "utils/emms.h" | ||
| 65 : | #include "utils/timer.h" | ||
| 66 : | Isibaar | 1.9 | #include "bitstream/mbcoding.h" |
| 67 : | Isibaar | 1.1 | |
| 68 : | suxen_drol | 1.33.2.22 | #if defined(ARCH_X86) |
| 69 : | suxen_drol | 1.31 | |
| 70 : | suxen_drol | 1.33.2.22 | #if defined(_MSC_VER) |
| 71 : | # include <windows.h> | ||
| 72 : | suxen_drol | 1.31 | #else |
| 73 : | suxen_drol | 1.33.2.22 | # include <signal.h> |
| 74 : | # include <setjmp.h> | ||
| 75 : | suxen_drol | 1.31 | |
| 76 : | suxen_drol | 1.33.2.22 | static jmp_buf mark; |
| 77 : | suxen_drol | 1.31 | |
| 78 : | suxen_drol | 1.33.2.22 | static void |
| 79 : | sigill_handler(int signal) | ||
| 80 : | { | ||
| 81 : | longjmp(mark, 1); | ||
| 82 : | } | ||
| 83 : | suxen_drol | 1.31 | #endif |
| 84 : | |||
| 85 : | |||
| 86 : | /* | ||
| 87 : | calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled | ||
| 88 : | return values: | ||
| 89 : | -1 : could not determine | ||
| 90 : | 0 : SIGILL was *not* signalled | ||
| 91 : | 1 : SIGILL was signalled | ||
| 92 : | */ | ||
| 93 : | |||
| 94 : | int | ||
| 95 : | sigill_check(void (*func)()) | ||
| 96 : | { | ||
| 97 : | suxen_drol | 1.33.2.22 | #if defined(_MSC_VER) |
| 98 : | suxen_drol | 1.31 | _try { |
| 99 : | func(); | ||
| 100 : | } | ||
| 101 : | _except(EXCEPTION_EXECUTE_HANDLER) { | ||
| 102 : | |||
| 103 : | if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) | ||
| 104 : | return 1; | ||
| 105 : | } | ||
| 106 : | return 0; | ||
| 107 : | #else | ||
| 108 : | void * old_handler; | ||
| 109 : | int jmpret; | ||
| 110 : | |||
| 111 : | |||
| 112 : | old_handler = signal(SIGILL, sigill_handler); | ||
| 113 : | if (old_handler == SIG_ERR) | ||
| 114 : | { | ||
| 115 : | return -1; | ||
| 116 : | } | ||
| 117 : | |||
| 118 : | jmpret = setjmp(mark); | ||
| 119 : | if (jmpret == 0) | ||
| 120 : | { | ||
| 121 : | func(); | ||
| 122 : | } | ||
| 123 : | |||
| 124 : | signal(SIGILL, old_handler); | ||
| 125 : | |||
| 126 : | return jmpret; | ||
| 127 : | #endif | ||
| 128 : | } | ||
| 129 : | #endif | ||
| 130 : | |||
| 131 : | suxen_drol | 1.33.2.22 | |
| 132 : | /* detect cpu flags */ | ||
| 133 : | static unsigned int | ||
| 134 : | detect_cpu_flags() | ||
| 135 : | { | ||
| 136 : | /* enable native assembly optimizations by default */ | ||
| 137 : | unsigned int cpu_flags = XVID_CPU_ASM; | ||
| 138 : | |||
| 139 : | #if defined(ARCH_X86) | ||
| 140 : | cpu_flags |= check_cpu_features(); | ||
| 141 : | if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) | ||
| 142 : | cpu_flags &= ~XVID_CPU_SSE; | ||
| 143 : | |||
| 144 : | if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) | ||
| 145 : | cpu_flags &= ~XVID_CPU_SSE2; | ||
| 146 : | #endif | ||
| 147 : | |||
| 148 : | #if defined(ARCH_PPC) | ||
| 149 : | #if defined(ARCH_PPC_ALTIVEC) | ||
| 150 : | cpu_flags |= XVID_CPU_ALTIVEC; | ||
| 151 : | #endif | ||
| 152 : | #endif | ||
| 153 : | |||
| 154 : | return cpu_flags; | ||
| 155 : | } | ||
| 156 : | |||
| 157 : | |||
| 158 : | edgomez | 1.16 | /***************************************************************************** |
| 159 : | * XviD Init Entry point | ||
| 160 : | * | ||
| 161 : | * Well this function initialize all internal function pointers according | ||
| 162 : | * to the CPU features forced by the library client or autodetected (depending | ||
| 163 : | * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all | ||
| 164 : | * image colorspace transformation tables. | ||
| 165 : | * | ||
| 166 : | * Returned value : XVID_ERR_OK | ||
| 167 : | * + API_VERSION in the input XVID_INIT_PARAM structure | ||
| 168 : | * + core build " " " " " | ||
| 169 : | * | ||
| 170 : | ****************************************************************************/ | ||
| 171 : | |||
| 172 : | suxen_drol | 1.33.2.13 | |
| 173 : | static | ||
| 174 : | int xvid_init_init(XVID_INIT_PARAM * init_param) | ||
| 175 : | Isibaar | 1.1 | { |
| 176 : | int cpu_flags; | ||
| 177 : | |||
| 178 : | suxen_drol | 1.21 | /* Inform the client the API version */ |
| 179 : | init_param->api_version = API_VERSION; | ||
| 180 : | |||
| 181 : | /* Inform the client the core build - unused because we're still alpha */ | ||
| 182 : | init_param->core_build = 1000; | ||
| 183 : | |||
| 184 : | suxen_drol | 1.31 | /* Do we have to force CPU features ? */ |
| 185 : | if ((init_param->cpu_flags & XVID_CPU_FORCE)) { | ||
| 186 : | suxen_drol | 1.21 | |
| 187 : | Isibaar | 1.1 | cpu_flags = init_param->cpu_flags; |
| 188 : | suxen_drol | 1.31 | |
| 189 : | edgomez | 1.16 | } else { |
| 190 : | Isibaar | 1.1 | |
| 191 : | suxen_drol | 1.33.2.22 | cpu_flags = detect_cpu_flags(); |
| 192 : | suxen_drol | 1.31 | } |
| 193 : | |||
| 194 : | if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) | ||
| 195 : | { | ||
| 196 : | Isibaar | 1.1 | init_param->cpu_flags = cpu_flags; |
| 197 : | suxen_drol | 1.31 | return XVID_ERR_OK; |
| 198 : | Isibaar | 1.1 | } |
| 199 : | suxen_drol | 1.31 | |
| 200 : | init_param->cpu_flags = cpu_flags; | ||
| 201 : | |||
| 202 : | Isibaar | 1.1 | |
| 203 : | edgomez | 1.16 | /* Initialize the function pointers */ |
| 204 : | Isibaar | 1.1 | idct_int32_init(); |
| 205 : | Isibaar | 1.9 | init_vlc_tables(); |
| 206 : | |||
| 207 : | edgomez | 1.16 | /* Fixed Point Forward/Inverse DCT transformations */ |
| 208 : | Isibaar | 1.1 | fdct = fdct_int32; |
| 209 : | idct = idct_int32; | ||
| 210 : | |||
| 211 : | edgomez | 1.16 | /* Only needed on PPC Altivec archs */ |
| 212 : | canard | 1.10 | sadInit = 0; |
| 213 : | edgomez | 1.15 | |
| 214 : | edgomez | 1.16 | /* Restore FPU context : emms_c is a nop functions */ |
| 215 : | Isibaar | 1.1 | emms = emms_c; |
| 216 : | |||
| 217 : | edgomez | 1.16 | /* Quantization functions */ |
| 218 : | quant_intra = quant_intra_c; | ||
| 219 : | Isibaar | 1.1 | dequant_intra = dequant_intra_c; |
| 220 : | edgomez | 1.16 | quant_inter = quant_inter_c; |
| 221 : | Isibaar | 1.1 | dequant_inter = dequant_inter_c; |
| 222 : | |||
| 223 : | edgomez | 1.16 | quant4_intra = quant4_intra_c; |
| 224 : | Isibaar | 1.1 | dequant4_intra = dequant4_intra_c; |
| 225 : | edgomez | 1.16 | quant4_inter = quant4_inter_c; |
| 226 : | Isibaar | 1.1 | dequant4_inter = dequant4_inter_c; |
| 227 : | |||
| 228 : | edgomez | 1.16 | /* Block transfer related functions */ |
| 229 : | Isibaar | 1.1 | transfer_8to16copy = transfer_8to16copy_c; |
| 230 : | transfer_16to8copy = transfer_16to8copy_c; | ||
| 231 : | edgomez | 1.16 | transfer_8to16sub = transfer_8to16sub_c; |
| 232 : | syskin | 1.33.2.23 | transfer_8to16subro = transfer_8to16subro_c; |
| 233 : | suxen_drol | 1.11 | transfer_8to16sub2 = transfer_8to16sub2_c; |
| 234 : | edgomez | 1.16 | transfer_16to8add = transfer_16to8add_c; |
| 235 : | transfer8x8_copy = transfer8x8_copy_c; | ||
| 236 : | Isibaar | 1.1 | |
| 237 : | h | 1.33.2.5 | /* Interlacing functions */ |
| 238 : | MBFieldTest = MBFieldTest_c; | ||
| 239 : | |||
| 240 : | edgomez | 1.16 | /* Image interpolation related functions */ |
| 241 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; | ||
| 242 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; | ||
| 243 : | Isibaar | 1.1 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
| 244 : | Isibaar | 1.33.2.8 | |
| 245 : | interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; | ||
| 246 : | interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; | ||
| 247 : | interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; | ||
| 248 : | Isibaar | 1.1 | |
| 249 : | Isibaar | 1.33.2.6 | interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; |
| 250 : | interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; | ||
| 251 : | interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; | ||
| 252 : | |||
| 253 : | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; | ||
| 254 : | interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; | ||
| 255 : | |||
| 256 : | interpolate8x8_avg2 = interpolate8x8_avg2_c; | ||
| 257 : | interpolate8x8_avg4 = interpolate8x8_avg4_c; | ||
| 258 : | |||
| 259 : | suxen_drol | 1.33.2.15 | /* reduced resoltuion */ |
| 260 : | suxen_drol | 1.33.2.17 | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
| 261 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; | ||
| 262 : | suxen_drol | 1.33.2.15 | vfilter_31 = xvid_VFilter_31_C; |
| 263 : | hfilter_31 = xvid_HFilter_31_C; | ||
| 264 : | suxen_drol | 1.33.2.17 | filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
| 265 : | filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; | ||
| 266 : | suxen_drol | 1.33.2.15 | |
| 267 : | edgomez | 1.16 | /* Initialize internal colorspace transformation tables */ |
| 268 : | Isibaar | 1.1 | colorspace_init(); |
| 269 : | |||
| 270 : | edgomez | 1.16 | /* All colorspace transformation functions User Format->YV12 */ |
| 271 : | suxen_drol | 1.33.2.11 | yv12_to_yv12 = yv12_to_yv12_c; |
| 272 : | rgb555_to_yv12 = rgb555_to_yv12_c; | ||
| 273 : | rgb565_to_yv12 = rgb565_to_yv12_c; | ||
| 274 : | bgr_to_yv12 = bgr_to_yv12_c; | ||
| 275 : | bgra_to_yv12 = bgra_to_yv12_c; | ||
| 276 : | abgr_to_yv12 = abgr_to_yv12_c; | ||
| 277 : | rgba_to_yv12 = rgba_to_yv12_c; | ||
| 278 : | yuyv_to_yv12 = yuyv_to_yv12_c; | ||
| 279 : | uyvy_to_yv12 = uyvy_to_yv12_c; | ||
| 280 : | |||
| 281 : | rgb555i_to_yv12 = rgb555i_to_yv12_c; | ||
| 282 : | rgb565i_to_yv12 = rgb565i_to_yv12_c; | ||
| 283 : | bgri_to_yv12 = bgri_to_yv12_c; | ||
| 284 : | bgrai_to_yv12 = bgrai_to_yv12_c; | ||
| 285 : | abgri_to_yv12 = abgri_to_yv12_c; | ||
| 286 : | rgbai_to_yv12 = rgbai_to_yv12_c; | ||
| 287 : | yuyvi_to_yv12 = yuyvi_to_yv12_c; | ||
| 288 : | uyvyi_to_yv12 = uyvyi_to_yv12_c; | ||
| 289 : | |||
| 290 : | Isibaar | 1.1 | |
| 291 : | edgomez | 1.16 | /* All colorspace transformation functions YV12->User format */ |
| 292 : | suxen_drol | 1.33.2.11 | yv12_to_rgb555 = yv12_to_rgb555_c; |
| 293 : | yv12_to_rgb565 = yv12_to_rgb565_c; | ||
| 294 : | yv12_to_bgr = yv12_to_bgr_c; | ||
| 295 : | yv12_to_bgra = yv12_to_bgra_c; | ||
| 296 : | yv12_to_abgr = yv12_to_abgr_c; | ||
| 297 : | yv12_to_rgba = yv12_to_rgba_c; | ||
| 298 : | yv12_to_yuyv = yv12_to_yuyv_c; | ||
| 299 : | yv12_to_uyvy = yv12_to_uyvy_c; | ||
| 300 : | |||
| 301 : | yv12_to_rgb555i = yv12_to_rgb555i_c; | ||
| 302 : | yv12_to_rgb565i = yv12_to_rgb565i_c; | ||
| 303 : | yv12_to_bgri = yv12_to_bgri_c; | ||
| 304 : | yv12_to_bgrai = yv12_to_bgrai_c; | ||
| 305 : | yv12_to_abgri = yv12_to_abgri_c; | ||
| 306 : | yv12_to_rgbai = yv12_to_rgbai_c; | ||
| 307 : | yv12_to_yuyvi = yv12_to_yuyvi_c; | ||
| 308 : | yv12_to_uyvyi = yv12_to_uyvyi_c; | ||
| 309 : | Isibaar | 1.1 | |
| 310 : | edgomez | 1.16 | /* Functions used in motion estimation algorithms */ |
| 311 : | Isibaar | 1.1 | calc_cbp = calc_cbp_c; |
| 312 : | edgomez | 1.16 | sad16 = sad16_c; |
| 313 : | suxen_drol | 1.33 | sad8 = sad8_c; |
| 314 : | edgomez | 1.16 | sad16bi = sad16bi_c; |
| 315 : | suxen_drol | 1.33 | sad8bi = sad8bi_c; |
| 316 : | edgomez | 1.16 | dev16 = dev16_c; |
| 317 : | chl | 1.33.2.1 | sad16v = sad16v_c; |
| 318 : | suxen_drol | 1.33 | |
| 319 : | chl | 1.33.2.1 | // Halfpel8_Refine = Halfpel8_Refine_c; |
| 320 : | Isibaar | 1.1 | |
| 321 : | suxen_drol | 1.33.2.22 | #if defined(ARCH_X86) |
| 322 : | |||
| 323 : | if ((cpu_flags & XVID_CPU_ASM)) | ||
| 324 : | { | ||
| 325 : | vfilter_31 = xvid_VFilter_31_x86; | ||
| 326 : | hfilter_31 = xvid_HFilter_31_x86; | ||
| 327 : | } | ||
| 328 : | suxen_drol | 1.33.2.11 | |
| 329 : | if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || | ||
| 330 : | (cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || | ||
| 331 : | (cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) | ||
| 332 : | { | ||
| 333 : | /* Restore FPU context : emms_c is a nop functions */ | ||
| 334 : | emms = emms_mmx; | ||
| 335 : | } | ||
| 336 : | |||
| 337 : | suxen_drol | 1.33.2.22 | if ((cpu_flags & XVID_CPU_MMX)) { |
| 338 : | edgomez | 1.16 | |
| 339 : | /* Forward and Inverse Discrete Cosine Transformation functions */ | ||
| 340 : | Isibaar | 1.1 | fdct = fdct_mmx; |
| 341 : | idct = idct_mmx; | ||
| 342 : | |||
| 343 : | edgomez | 1.16 | /* Quantization related functions */ |
| 344 : | quant_intra = quant_intra_mmx; | ||
| 345 : | Isibaar | 1.1 | dequant_intra = dequant_intra_mmx; |
| 346 : | edgomez | 1.16 | quant_inter = quant_inter_mmx; |
| 347 : | Isibaar | 1.1 | dequant_inter = dequant_inter_mmx; |
| 348 : | |||
| 349 : | edgomez | 1.16 | quant4_intra = quant4_intra_mmx; |
| 350 : | Isibaar | 1.1 | dequant4_intra = dequant4_intra_mmx; |
| 351 : | edgomez | 1.16 | quant4_inter = quant4_inter_mmx; |
| 352 : | Isibaar | 1.1 | dequant4_inter = dequant4_inter_mmx; |
| 353 : | |||
| 354 : | edgomez | 1.16 | /* Block related functions */ |
| 355 : | Isibaar | 1.1 | transfer_8to16copy = transfer_8to16copy_mmx; |
| 356 : | transfer_16to8copy = transfer_16to8copy_mmx; | ||
| 357 : | edgomez | 1.16 | transfer_8to16sub = transfer_8to16sub_mmx; |
| 358 : | syskin | 1.33.2.23 | transfer_8to16subro = transfer_8to16subro_mmx; |
| 359 : | edgomez | 1.22 | transfer_8to16sub2 = transfer_8to16sub2_mmx; |
| 360 : | edgomez | 1.16 | transfer_16to8add = transfer_16to8add_mmx; |
| 361 : | transfer8x8_copy = transfer8x8_copy_mmx; | ||
| 362 : | edgomez | 1.22 | |
| 363 : | h | 1.33.2.5 | /* Interlacing Functions */ |
| 364 : | MBFieldTest = MBFieldTest_mmx; | ||
| 365 : | edgomez | 1.16 | |
| 366 : | /* Image Interpolation related functions */ | ||
| 367 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; | ||
| 368 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; | ||
| 369 : | Isibaar | 1.1 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
| 370 : | Isibaar | 1.33.2.6 | |
| 371 : | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; | ||
| 372 : | interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; | ||
| 373 : | |||
| 374 : | Isibaar | 1.33.2.14 | interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
| 375 : | Isibaar | 1.33.2.6 | interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
| 376 : | Isibaar | 1.1 | |
| 377 : | suxen_drol | 1.33.2.15 | /* reduced resolution */ |
| 378 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; | ||
| 379 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; | ||
| 380 : | hfilter_31 = xvid_HFilter_31_mmx; | ||
| 381 : | suxen_drol | 1.33.2.17 | filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; |
| 382 : | filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; | ||
| 383 : | suxen_drol | 1.33.2.15 | |
| 384 : | suxen_drol | 1.33.2.11 | /* image input xxx_to_yv12 related functions */ |
| 385 : | yv12_to_yv12 = yv12_to_yv12_mmx; | ||
| 386 : | bgr_to_yv12 = bgr_to_yv12_mmx; | ||
| 387 : | bgra_to_yv12 = bgra_to_yv12_mmx; | ||
| 388 : | edgomez | 1.16 | yuyv_to_yv12 = yuyv_to_yv12_mmx; |
| 389 : | uyvy_to_yv12 = uyvy_to_yv12_mmx; | ||
| 390 : | Isibaar | 1.1 | |
| 391 : | suxen_drol | 1.33.2.11 | /* image output yv12_to_xxx related functions */ |
| 392 : | yv12_to_bgr = yv12_to_bgr_mmx; | ||
| 393 : | yv12_to_bgra = yv12_to_bgra_mmx; | ||
| 394 : | edgomez | 1.16 | yv12_to_yuyv = yv12_to_yuyv_mmx; |
| 395 : | yv12_to_uyvy = yv12_to_uyvy_mmx; | ||
| 396 : | suxen_drol | 1.33.2.11 | |
| 397 : | yv12_to_yuyvi = yv12_to_yuyvi_mmx; | ||
| 398 : | yv12_to_uyvyi = yv12_to_uyvyi_mmx; | ||
| 399 : | Isibaar | 1.1 | |
| 400 : | edgomez | 1.16 | /* Motion estimation related functions */ |
| 401 : | Isibaar | 1.1 | calc_cbp = calc_cbp_mmx; |
| 402 : | edgomez | 1.16 | sad16 = sad16_mmx; |
| 403 : | sad8 = sad8_mmx; | ||
| 404 : | suxen_drol | 1.33 | sad16bi = sad16bi_mmx; |
| 405 : | sad8bi = sad8bi_mmx; | ||
| 406 : | edgomez | 1.16 | dev16 = dev16_mmx; |
| 407 : | Isibaar | 1.33.2.2 | sad16v = sad16v_mmx; |
| 408 : | Isibaar | 1.1 | } |
| 409 : | |||
| 410 : | suxen_drol | 1.33 | /* these 3dnow functions are faster than mmx, but slower than xmm. */ |
| 411 : | suxen_drol | 1.33.2.22 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
| 412 : | |||
| 413 : | emms = emms_3dn; | ||
| 414 : | suxen_drol | 1.33 | |
| 415 : | /* ME functions */ | ||
| 416 : | sad16bi = sad16bi_3dn; | ||
| 417 : | sad8bi = sad8bi_3dn; | ||
| 418 : | suxen_drol | 1.33.2.11 | |
| 419 : | yuyv_to_yv12 = yuyv_to_yv12_3dn; | ||
| 420 : | uyvy_to_yv12 = uyvy_to_yv12_3dn; | ||
| 421 : | suxen_drol | 1.33 | } |
| 422 : | |||
| 423 : | |||
| 424 : | suxen_drol | 1.33.2.22 | if ((cpu_flags & XVID_CPU_MMXEXT)) { |
| 425 : | edgomez | 1.16 | |
| 426 : | /* Inverse DCT */ | ||
| 427 : | Isibaar | 1.1 | idct = idct_xmm; |
| 428 : | edgomez | 1.16 | |
| 429 : | /* Interpolation */ | ||
| 430 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; | ||
| 431 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; | ||
| 432 : | h | 1.3 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
| 433 : | suxen_drol | 1.33.2.15 | |
| 434 : | /* reduced resolution */ | ||
| 435 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; | ||
| 436 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; | ||
| 437 : | Isibaar | 1.25 | |
| 438 : | chenm001 | 1.29 | /* Quantization */ |
| 439 : | Isibaar | 1.33.2.18 | quant4_intra = quant4_intra_xmm; |
| 440 : | quant4_inter = quant4_inter_xmm; | ||
| 441 : | |||
| 442 : | chenm001 | 1.29 | dequant_intra = dequant_intra_xmm; |
| 443 : | dequant_inter = dequant_inter_xmm; | ||
| 444 : | |||
| 445 : | edgomez | 1.19 | /* Buffer transfer */ |
| 446 : | transfer_8to16sub2 = transfer_8to16sub2_xmm; | ||
| 447 : | edgomez | 1.16 | |
| 448 : | /* Colorspace transformation */ | ||
| 449 : | suxen_drol | 1.33.2.11 | yv12_to_yv12 = yv12_to_yv12_xmm; |
| 450 : | yuyv_to_yv12 = yuyv_to_yv12_xmm; | ||
| 451 : | Isibaar | 1.33.2.12 | uyvy_to_yv12 = uyvy_to_yv12_xmm; |
| 452 : | Isibaar | 1.1 | |
| 453 : | edgomez | 1.16 | /* ME functions */ |
| 454 : | Isibaar | 1.1 | sad16 = sad16_xmm; |
| 455 : | suxen_drol | 1.33 | sad8 = sad8_xmm; |
| 456 : | chenm001 | 1.29 | sad16bi = sad16bi_xmm; |
| 457 : | suxen_drol | 1.33 | sad8bi = sad8bi_xmm; |
| 458 : | Isibaar | 1.1 | dev16 = dev16_xmm; |
| 459 : | chl | 1.33.2.1 | sad16v = sad16v_xmm; |
| 460 : | Isibaar | 1.1 | } |
| 461 : | |||
| 462 : | suxen_drol | 1.33.2.22 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
| 463 : | edgomez | 1.16 | |
| 464 : | /* Interpolation */ | ||
| 465 : | Isibaar | 1.1 | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
| 466 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; | ||
| 467 : | h | 1.4 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
| 468 : | Isibaar | 1.1 | } |
| 469 : | Isibaar | 1.33.2.18 | |
| 470 : | suxen_drol | 1.33.2.22 | if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
| 471 : | Isibaar | 1.33.2.18 | |
| 472 : | /* Inverse DCT */ | ||
| 473 : | idct = idct_3dne; | ||
| 474 : | |||
| 475 : | /* Buffer transfer */ | ||
| 476 : | transfer_8to16copy = transfer_8to16copy_3dne; | ||
| 477 : | transfer_16to8copy = transfer_16to8copy_3dne; | ||
| 478 : | transfer_8to16sub = transfer_8to16sub_3dne; | ||
| 479 : | syskin | 1.33.2.23 | transfer_8to16subro = transfer_8to16subro_3dne; |
| 480 : | Isibaar | 1.33.2.18 | transfer_8to16sub2 = transfer_8to16sub2_3dne; |
| 481 : | transfer_16to8add = transfer_16to8add_3dne; | ||
| 482 : | transfer8x8_copy = transfer8x8_copy_3dne; | ||
| 483 : | |||
| 484 : | /* Quantization */ | ||
| 485 : | dequant4_intra = dequant4_intra_3dne; | ||
| 486 : | dequant4_inter = dequant4_inter_3dne; | ||
| 487 : | quant_intra = quant_intra_3dne; | ||
| 488 : | quant_inter = quant_inter_3dne; | ||
| 489 : | dequant_intra = dequant_intra_3dne; | ||
| 490 : | dequant_inter = dequant_inter_3dne; | ||
| 491 : | |||
| 492 : | /* ME functions */ | ||
| 493 : | calc_cbp = calc_cbp_3dne; | ||
| 494 : | sad16 = sad16_3dne; | ||
| 495 : | sad8 = sad8_3dne; | ||
| 496 : | sad16bi = sad16bi_3dne; | ||
| 497 : | sad8bi = sad8bi_3dne; | ||
| 498 : | dev16 = dev16_3dne; | ||
| 499 : | |||
| 500 : | /* Interpolation */ | ||
| 501 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; | ||
| 502 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; | ||
| 503 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; | ||
| 504 : | } | ||
| 505 : | |||
| 506 : | Isibaar | 1.1 | |
| 507 : | suxen_drol | 1.33.2.22 | if ((cpu_flags & XVID_CPU_SSE2)) { |
| 508 : | edgomez | 1.16 | |
| 509 : | chenm001 | 1.29 | calc_cbp = calc_cbp_sse2; |
| 510 : | |||
| 511 : | edgomez | 1.16 | /* Quantization */ |
| 512 : | quant_intra = quant_intra_sse2; | ||
| 513 : | Isibaar | 1.14 | dequant_intra = dequant_intra_sse2; |
| 514 : | edgomez | 1.16 | quant_inter = quant_inter_sse2; |
| 515 : | Isibaar | 1.14 | dequant_inter = dequant_inter_sse2; |
| 516 : | h | 1.13 | |
| 517 : | suxen_drol | 1.33.2.22 | #if defined(EXPERIMENTAL_SSE2_CODE) |
| 518 : | /* ME; slower than xmm */ | ||
| 519 : | edgomez | 1.16 | sad16 = sad16_sse2; |
| 520 : | dev16 = dev16_sse2; | ||
| 521 : | suxen_drol | 1.33.2.22 | #endif |
| 522 : | edgomez | 1.16 | /* Forward and Inverse DCT */ |
| 523 : | idct = idct_sse2; | ||
| 524 : | Isibaar | 1.14 | fdct = fdct_sse2; |
| 525 : | h | 1.12 | } |
| 526 : | Isibaar | 1.1 | #endif |
| 527 : | Isibaar | 1.18 | |
| 528 : | suxen_drol | 1.33.2.22 | #if defined(ARCH_IA64) |
| 529 : | if ((cpu_flags & XVID_CPU_ASM)) { //use assembler routines? | ||
| 530 : | Isibaar | 1.18 | idct_ia64_init(); |
| 531 : | fdct = fdct_ia64; | ||
| 532 : | idct = idct_ia64; //not yet working, crashes | ||
| 533 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; | ||
| 534 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; | ||
| 535 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; | ||
| 536 : | sad16 = sad16_ia64; | ||
| 537 : | sad16bi = sad16bi_ia64; | ||
| 538 : | sad8 = sad8_ia64; | ||
| 539 : | dev16 = dev16_ia64; | ||
| 540 : | chl | 1.33.2.1 | // Halfpel8_Refine = Halfpel8_Refine_ia64; |
| 541 : | Isibaar | 1.18 | quant_intra = quant_intra_ia64; |
| 542 : | dequant_intra = dequant_intra_ia64; | ||
| 543 : | quant_inter = quant_inter_ia64; | ||
| 544 : | dequant_inter = dequant_inter_ia64; | ||
| 545 : | transfer_8to16copy = transfer_8to16copy_ia64; | ||
| 546 : | transfer_16to8copy = transfer_16to8copy_ia64; | ||
| 547 : | transfer_8to16sub = transfer_8to16sub_ia64; | ||
| 548 : | transfer_8to16sub2 = transfer_8to16sub2_ia64; | ||
| 549 : | transfer_16to8add = transfer_16to8add_ia64; | ||
| 550 : | transfer8x8_copy = transfer8x8_copy_ia64; | ||
| 551 : | DEBUG("Using IA-64 assembler routines.\n"); | ||
| 552 : | } | ||
| 553 : | #endif | ||
| 554 : | edgomez | 1.16 | |
| 555 : | suxen_drol | 1.33.2.22 | #if defined(ARCH_PPC) |
| 556 : | if ((cpu_flags & XVID_CPU_ASM)) | ||
| 557 : | { | ||
| 558 : | calc_cbp = calc_cbp_ppc; | ||
| 559 : | } | ||
| 560 : | |||
| 561 : | if ((cpu_flags & XVID_CPU_ALTIVEC)) | ||
| 562 : | { | ||
| 563 : | calc_cbp = calc_cbp_altivec; | ||
| 564 : | fdct = fdct_altivec; | ||
| 565 : | idct = idct_altivec; | ||
| 566 : | sadInit = sadInit_altivec; | ||
| 567 : | sad16 = sad16_altivec; | ||
| 568 : | sad8 = sad8_altivec; | ||
| 569 : | dev16 = dev16_altivec; | ||
| 570 : | } | ||
| 571 : | canard | 1.5 | #endif |
| 572 : | Isibaar | 1.1 | |
| 573 : | return XVID_ERR_OK; | ||
| 574 : | suxen_drol | 1.33.2.13 | } |
| 575 : | |||
| 576 : | |||
| 577 : | |||
| 578 : | static int | ||
| 579 : | xvid_init_convert(XVID_INIT_CONVERTINFO* convert) | ||
| 580 : | { | ||
| 581 : | suxen_drol | 1.33.2.16 | // const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
| 582 : | suxen_drol | 1.33.2.13 | const int width = convert->width; |
| 583 : | const int height = convert->height; | ||
| 584 : | const int width2 = convert->width/2; | ||
| 585 : | const int height2 = convert->height/2; | ||
| 586 : | IMAGE img; | ||
| 587 : | |||
| 588 : | switch (convert->input.colorspace & ~XVID_CSP_VFLIP) | ||
| 589 : | { | ||
| 590 : | case XVID_CSP_YV12 : | ||
| 591 : | img.y = convert->input.y; | ||
| 592 : | img.v = (uint8_t*)convert->input.y + width*height; | ||
| 593 : | img.u = (uint8_t*)convert->input.y + width*height + width2*height2; | ||
| 594 : | image_output(&img, width, height, width, | ||
| 595 : | convert->output.y, convert->output.y_stride, | ||
| 596 : | convert->output.colorspace, convert->interlacing); | ||
| 597 : | break; | ||
| 598 : | |||
| 599 : | default : | ||
| 600 : | return XVID_ERR_FORMAT; | ||
| 601 : | } | ||
| 602 : | |||
| 603 : | |||
| 604 : | emms(); | ||
| 605 : | return XVID_ERR_OK; | ||
| 606 : | } | ||
| 607 : | |||
| 608 : | |||
| 609 : | suxen_drol | 1.33.2.19 | |
| 610 : | void fill8(uint8_t * block, int size, int value) | ||
| 611 : | { | ||
| 612 : | int i; | ||
| 613 : | for (i = 0; i < size; i++) | ||
| 614 : | block[i] = value; | ||
| 615 : | } | ||
| 616 : | |||
| 617 : | void fill16(int16_t * block, int size, int value) | ||
| 618 : | { | ||
| 619 : | int i; | ||
| 620 : | for (i = 0; i < size; i++) | ||
| 621 : | block[i] = value; | ||
| 622 : | } | ||
| 623 : | |||
| 624 : | #define RANDOM(min,max) min + (rand() % (max-min)) | ||
| 625 : | |||
| 626 : | void random8(uint8_t * block, int size, int min, int max) | ||
| 627 : | { | ||
| 628 : | int i; | ||
| 629 : | for (i = 0; i < size; i++) | ||
| 630 : | block[i] = RANDOM(min,max); | ||
| 631 : | } | ||
| 632 : | |||
| 633 : | void random16(int16_t * block, int size, int min, int max) | ||
| 634 : | { | ||
| 635 : | int i; | ||
| 636 : | for (i = 0; i < size; i++) | ||
| 637 : | block[i] = RANDOM(min,max); | ||
| 638 : | } | ||
| 639 : | |||
| 640 : | int compare16(const int16_t * blockA, const int16_t * blockB, int size) | ||
| 641 : | { | ||
| 642 : | int i; | ||
| 643 : | for (i = 0; i < size; i++) | ||
| 644 : | if (blockA[i] != blockB[i]) | ||
| 645 : | return 1; | ||
| 646 : | |||
| 647 : | return 0; | ||
| 648 : | } | ||
| 649 : | |||
| 650 : | suxen_drol | 1.33.2.20 | int diff16(const int16_t * blockA, const int16_t * blockB, int size) |
| 651 : | { | ||
| 652 : | int i, diff = 0; | ||
| 653 : | for (i = 0; i < size; i++) | ||
| 654 : | diff += ABS(blockA[i]-blockB[i]); | ||
| 655 : | return diff; | ||
| 656 : | } | ||
| 657 : | |||
| 658 : | |||
| 659 : | #define XVID_TEST_RANDOM 0x00000001 /* random input data */ | ||
| 660 : | #define XVID_TEST_VERBOSE 0x00000002 /* verbose error output */ | ||
| 661 : | suxen_drol | 1.33.2.19 | |
| 662 : | |||
| 663 : | suxen_drol | 1.33.2.20 | #define TEST_FORWARD 0x00000001 /* intra */ |
| 664 : | #define TEST_FDCT (TEST_FORWARD) | ||
| 665 : | #define TEST_IDCT (0) | ||
| 666 : | |||
| 667 : | int test_transform(void * funcA, void * funcB, const char * nameB, | ||
| 668 : | int test, int flags) | ||
| 669 : | suxen_drol | 1.33.2.19 | { |
| 670 : | suxen_drol | 1.33.2.20 | int i; |
| 671 : | suxen_drol | 1.33.2.19 | int64_t timeSTART; |
| 672 : | int64_t timeA = 0; | ||
| 673 : | int64_t timeB = 0; | ||
| 674 : | DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); | ||
| 675 : | DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); | ||
| 676 : | suxen_drol | 1.33.2.20 | int min, max; |
| 677 : | int count = 0; | ||
| 678 : | suxen_drol | 1.33.2.19 | |
| 679 : | suxen_drol | 1.33.2.20 | int tmp; |
| 680 : | int min_error = 0x10000*64; | ||
| 681 : | int max_error = 0; | ||
| 682 : | |||
| 683 : | |||
| 684 : | if ((test & TEST_FORWARD)) /* forward */ | ||
| 685 : | suxen_drol | 1.33.2.19 | { |
| 686 : | suxen_drol | 1.33.2.20 | min = -256; |
| 687 : | max = 255; | ||
| 688 : | }else{ /* inverse */ | ||
| 689 : | min = -2048; | ||
| 690 : | max = 2047; | ||
| 691 : | } | ||
| 692 : | |||
| 693 : | for (i = 0; i < 64*64; i++) | ||
| 694 : | { | ||
| 695 : | if ((flags & XVID_TEST_RANDOM)) | ||
| 696 : | suxen_drol | 1.33.2.19 | { |
| 697 : | suxen_drol | 1.33.2.20 | random16(arrayA, 64, min, max); |
| 698 : | }else{ | ||
| 699 : | fill16(arrayA, 64, i); | ||
| 700 : | } | ||
| 701 : | memcpy(arrayB, arrayA, 64*sizeof(int16_t)); | ||
| 702 : | suxen_drol | 1.33.2.19 | |
| 703 : | suxen_drol | 1.33.2.20 | if ((test & TEST_FORWARD)) |
| 704 : | { | ||
| 705 : | suxen_drol | 1.33.2.19 | timeSTART = read_counter(); |
| 706 : | suxen_drol | 1.33.2.20 | ((fdctFunc*)funcA)(arrayA); |
| 707 : | suxen_drol | 1.33.2.19 | timeA += read_counter() - timeSTART; |
| 708 : | |||
| 709 : | timeSTART = read_counter(); | ||
| 710 : | suxen_drol | 1.33.2.20 | ((fdctFunc*)funcB)(arrayB); |
| 711 : | suxen_drol | 1.33.2.19 | timeB += read_counter() - timeSTART; |
| 712 : | suxen_drol | 1.33.2.20 | } |
| 713 : | else | ||
| 714 : | { | ||
| 715 : | timeSTART = read_counter(); | ||
| 716 : | ((idctFunc*)funcA)(arrayA); | ||
| 717 : | timeA += read_counter() - timeSTART; | ||
| 718 : | suxen_drol | 1.33.2.19 | |
| 719 : | suxen_drol | 1.33.2.20 | timeSTART = read_counter(); |
| 720 : | ((idctFunc*)funcB)(arrayB); | ||
| 721 : | timeB += read_counter() - timeSTART; | ||
| 722 : | suxen_drol | 1.33.2.19 | } |
| 723 : | suxen_drol | 1.33.2.20 | |
| 724 : | tmp = diff16(arrayA, arrayB, 64) / 64; | ||
| 725 : | if (tmp > max_error) | ||
| 726 : | max_error = tmp; | ||
| 727 : | if (tmp < min_error) | ||
| 728 : | min_error = tmp; | ||
| 729 : | |||
| 730 : | count++; | ||
| 731 : | suxen_drol | 1.33.2.19 | } |
| 732 : | |||
| 733 : | suxen_drol | 1.33.2.20 | /* print the "average difference" of best/worst transforms */ |
| 734 : | suxen_drol | 1.33.2.21 | printf("%s:\t%i\t(min_error:%i, max_error:%i)\n", nameB, (int)(timeB / count), min_error, max_error); |
| 735 : | suxen_drol | 1.33.2.19 | |
| 736 : | return 0; | ||
| 737 : | } | ||
| 738 : | |||
| 739 : | suxen_drol | 1.33.2.20 | |
| 740 : | #define TEST_QUANT 0x00000001 /* forward quantization */ | ||
| 741 : | #define TEST_INTRA 0x00000002 /* intra */ | ||
| 742 : | #define TEST_QUANT_INTRA (TEST_QUANT|TEST_INTRA) | ||
| 743 : | #define TEST_QUANT_INTER (TEST_QUANT) | ||
| 744 : | #define TEST_DEQUANT_INTRA (TEST_INTRA) | ||
| 745 : | #define TEST_DEQUANT_INTER (0) | ||
| 746 : | |||
| 747 : | int test_quant(void * funcA, void * funcB, const char * nameB, | ||
| 748 : | int test, int flags) | ||
| 749 : | suxen_drol | 1.33.2.19 | { |
| 750 : | int q,i; | ||
| 751 : | int64_t timeSTART; | ||
| 752 : | int64_t timeA = 0; | ||
| 753 : | int64_t timeB = 0; | ||
| 754 : | suxen_drol | 1.33.2.20 | int retA, retB; |
| 755 : | suxen_drol | 1.33.2.19 | DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
| 756 : | DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); | ||
| 757 : | DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); | ||
| 758 : | suxen_drol | 1.33.2.20 | int min, max; |
| 759 : | int count = 0; | ||
| 760 : | int errors = 0; | ||
| 761 : | |||
| 762 : | if ((test & TEST_QUANT)) /* quant */ | ||
| 763 : | { | ||
| 764 : | min = -2048; | ||
| 765 : | max = 2047; | ||
| 766 : | }else{ /* dequant */ | ||
| 767 : | min = -256; | ||
| 768 : | max = 255; | ||
| 769 : | } | ||
| 770 : | |||
| 771 : | suxen_drol | 1.33.2.19 | for (q = 1; q <= 31; q++) /* quantizer */ |
| 772 : | { | ||
| 773 : | for (i = min; i < max; i++) /* input coeff */ | ||
| 774 : | { | ||
| 775 : | suxen_drol | 1.33.2.20 | if ((flags & XVID_TEST_RANDOM)) |
| 776 : | { | ||
| 777 : | random16(arrayX, 64, min, max); | ||
| 778 : | }else{ | ||
| 779 : | fill16(arrayX, 64, i); | ||
| 780 : | } | ||
| 781 : | suxen_drol | 1.33.2.19 | |
| 782 : | suxen_drol | 1.33.2.20 | if ((test & TEST_INTRA)) /* intra */ |
| 783 : | { | ||
| 784 : | timeSTART = read_counter(); | ||
| 785 : | ((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q); | ||
| 786 : | timeA += read_counter() - timeSTART; | ||
| 787 : | |||
| 788 : | timeSTART = read_counter(); | ||
| 789 : | ((quanth263_intraFunc*)funcB)(arrayB, arrayX, q, q); | ||
| 790 : | timeB += read_counter() - timeSTART; | ||
| 791 : | } | ||
| 792 : | else /* inter */ | ||
| 793 : | { | ||
| 794 : | timeSTART = read_counter(); | ||
| 795 : | retA = ((quanth263_interFunc*)funcA)(arrayA, arrayX, q); | ||
| 796 : | timeA += read_counter() - timeSTART; | ||
| 797 : | |||
| 798 : | timeSTART = read_counter(); | ||
| 799 : | retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q); | ||
| 800 : | timeB += read_counter() - timeSTART; | ||
| 801 : | } | ||
| 802 : | suxen_drol | 1.33.2.19 | |
| 803 : | suxen_drol | 1.33.2.20 | /* compare return value from quant_inter, and compare (de)quantiz'd arrays */ |
| 804 : | if ( ((test&TEST_QUANT) && !(test&TEST_INTRA) && retA != retB ) || | ||
| 805 : | compare16(arrayA, arrayB, 64)) | ||
| 806 : | suxen_drol | 1.33.2.19 | { |
| 807 : | suxen_drol | 1.33.2.20 | errors++; |
| 808 : | if ((flags & XVID_TEST_VERBOSE)) | ||
| 809 : | printf("%s error: q=%i, i=%i\n", nameB, q, i); | ||
| 810 : | suxen_drol | 1.33.2.19 | } |
| 811 : | suxen_drol | 1.33.2.20 | |
| 812 : | count++; | ||
| 813 : | suxen_drol | 1.33.2.19 | } |
| 814 : | } | ||
| 815 : | |||
| 816 : | suxen_drol | 1.33.2.21 | printf("%s:\t%i", nameB, (int)(timeB / count)); |
| 817 : | suxen_drol | 1.33.2.20 | if (errors>0) |
| 818 : | printf("\t(%i errors out of %i)", errors, count); | ||
| 819 : | printf("\n"); | ||
| 820 : | |||
| 821 : | suxen_drol | 1.33.2.19 | return 0; |
| 822 : | } | ||
| 823 : | |||
| 824 : | |||
| 825 : | |||
| 826 : | suxen_drol | 1.33.2.20 | int xvid_init_test(int flags) |
| 827 : | suxen_drol | 1.33.2.19 | { |
| 828 : | int cpu_flags; | ||
| 829 : | |||
| 830 : | suxen_drol | 1.33.2.20 | srand(time(0)); |
| 831 : | |||
| 832 : | suxen_drol | 1.33.2.19 | printf("xvid_init_test\n"); |
| 833 : | |||
| 834 : | #if defined(ARCH_X86) | ||
| 835 : | suxen_drol | 1.33.2.22 | cpu_flags = detect_cpu_flags(); |
| 836 : | suxen_drol | 1.33.2.20 | idct_int32_init(); |
| 837 : | suxen_drol | 1.33.2.19 | emms_mmx(); |
| 838 : | |||
| 839 : | suxen_drol | 1.33.2.20 | printf("--- fdct ---\n"); |
| 840 : | test_transform(fdct_int32, fdct_int32, "c", TEST_FDCT, flags); | ||
| 841 : | if (cpu_flags & XVID_CPU_MMX) | ||
| 842 : | test_transform(fdct_int32, fdct_mmx, "mmx", TEST_FDCT, flags); | ||
| 843 : | if (cpu_flags & XVID_CPU_SSE2) | ||
| 844 : | test_transform(fdct_int32, fdct_sse2, "sse2", TEST_FDCT, flags); | ||
| 845 : | |||
| 846 : | printf("\n--- idct ---\n"); | ||
| 847 : | test_transform(idct_int32, idct_int32, "c", TEST_IDCT, flags); | ||
| 848 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 849 : | suxen_drol | 1.33.2.20 | test_transform(idct_int32, idct_mmx, "mmx", TEST_IDCT, flags); |
| 850 : | if (cpu_flags & XVID_CPU_MMXEXT) | ||
| 851 : | test_transform(idct_int32, idct_xmm, "xmm", TEST_IDCT, flags); | ||
| 852 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 853 : | suxen_drol | 1.33.2.20 | test_transform(idct_int32, idct_3dne, "3dne", TEST_IDCT, flags); |
| 854 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_SSE2) |
| 855 : | suxen_drol | 1.33.2.20 | test_transform(idct_int32, idct_sse2, "sse2", TEST_IDCT, flags); |
| 856 : | |||
| 857 : | printf("\n--- quant intra ---\n"); | ||
| 858 : | test_quant(quant_intra_c, quant_intra_c, "c", TEST_QUANT_INTRA, flags); | ||
| 859 : | if (cpu_flags & XVID_CPU_MMX) | ||
| 860 : | test_quant(quant_intra_c, quant_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); | ||
| 861 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
| 862 : | test_quant(quant_intra_c, quant_intra_3dne, "3dne", TEST_QUANT_INTRA, flags); | ||
| 863 : | if (cpu_flags & XVID_CPU_SSE2) | ||
| 864 : | test_quant(quant_intra_c, quant_intra_sse2, "sse2", TEST_QUANT_INTRA, flags); | ||
| 865 : | suxen_drol | 1.33.2.19 | |
| 866 : | printf("\n--- quant inter ---\n"); | ||
| 867 : | suxen_drol | 1.33.2.20 | test_quant(quant_inter_c, quant_inter_c, "c", TEST_QUANT_INTER, flags); |
| 868 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 869 : | suxen_drol | 1.33.2.20 | test_quant(quant_inter_c, quant_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
| 870 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 871 : | suxen_drol | 1.33.2.20 | test_quant(quant_inter_c, quant_inter_3dne, "3dne", TEST_QUANT_INTER, flags); |
| 872 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_SSE2) |
| 873 : | suxen_drol | 1.33.2.20 | test_quant(quant_inter_c, quant_inter_sse2, "sse2", TEST_QUANT_INTER, flags); |
| 874 : | suxen_drol | 1.33.2.19 | |
| 875 : | suxen_drol | 1.33.2.20 | printf("\n--- dequant intra ---\n"); |
| 876 : | test_quant(dequant_intra_c, dequant_intra_c, "c", TEST_DEQUANT_INTRA, flags); | ||
| 877 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 878 : | suxen_drol | 1.33.2.20 | test_quant(dequant_intra_c, dequant_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
| 879 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMXEXT) |
| 880 : | suxen_drol | 1.33.2.20 | test_quant(dequant_intra_c, dequant_intra_xmm, "xmm", TEST_DEQUANT_INTRA, flags); |
| 881 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 882 : | suxen_drol | 1.33.2.20 | test_quant(dequant_intra_c, dequant_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
| 883 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_SSE2) |
| 884 : | suxen_drol | 1.33.2.20 | test_quant(dequant_intra_c, dequant_intra_sse2, "sse2", TEST_DEQUANT_INTRA, flags); |
| 885 : | suxen_drol | 1.33.2.19 | |
| 886 : | printf("\n--- dequant inter ---\n"); | ||
| 887 : | suxen_drol | 1.33.2.20 | test_quant(dequant_inter_c, dequant_inter_c, "c", TEST_DEQUANT_INTER, flags); |
| 888 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 889 : | suxen_drol | 1.33.2.20 | test_quant(dequant_inter_c, dequant_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
| 890 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMXEXT) |
| 891 : | suxen_drol | 1.33.2.20 | test_quant(dequant_inter_c, dequant_inter_xmm, "xmm", TEST_DEQUANT_INTER, flags); |
| 892 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 893 : | suxen_drol | 1.33.2.20 | test_quant(dequant_inter_c, dequant_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
| 894 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_SSE2) |
| 895 : | suxen_drol | 1.33.2.20 | test_quant(dequant_inter_c, dequant_inter_sse2, "sse2", TEST_DEQUANT_INTER, flags); |
| 896 : | suxen_drol | 1.33.2.19 | |
| 897 : | printf("\n--- quant4_intra ---\n"); | ||
| 898 : | suxen_drol | 1.33.2.20 | test_quant(quant4_intra_c, quant4_intra_c, "c", TEST_QUANT_INTRA, flags); |
| 899 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 900 : | suxen_drol | 1.33.2.20 | test_quant(quant4_intra_c, quant4_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); |
| 901 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMXEXT) |
| 902 : | suxen_drol | 1.33.2.20 | test_quant(quant4_intra_c, quant4_intra_xmm, "xmm", TEST_QUANT_INTRA, flags); |
| 903 : | suxen_drol | 1.33.2.19 | |
| 904 : | printf("\n--- quant4_inter ---\n"); | ||
| 905 : | suxen_drol | 1.33.2.20 | test_quant(quant4_inter_c, quant4_inter_c, "c", TEST_QUANT_INTER, flags); |
| 906 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 907 : | suxen_drol | 1.33.2.20 | test_quant(quant4_inter_c, quant4_inter_mmx, "mmx", TEST_QUANT_INTER, flags); |
| 908 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMXEXT) |
| 909 : | suxen_drol | 1.33.2.20 | test_quant(quant4_inter_c, quant4_inter_xmm, "xmm", TEST_QUANT_INTER, flags); |
| 910 : | suxen_drol | 1.33.2.19 | |
| 911 : | printf("\n--- dequant4_intra ---\n"); | ||
| 912 : | suxen_drol | 1.33.2.20 | test_quant(dequant4_intra_c, dequant4_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
| 913 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 914 : | suxen_drol | 1.33.2.20 | test_quant(dequant4_intra_c, dequant4_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); |
| 915 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 916 : | suxen_drol | 1.33.2.20 | test_quant(dequant4_intra_c, dequant4_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); |
| 917 : | suxen_drol | 1.33.2.19 | |
| 918 : | printf("\n--- dequant4_inter ---\n"); | ||
| 919 : | suxen_drol | 1.33.2.20 | test_quant(dequant4_inter_c, dequant4_inter_c, "c", TEST_DEQUANT_INTER, flags); |
| 920 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_MMX) |
| 921 : | suxen_drol | 1.33.2.20 | test_quant(dequant4_inter_c, dequant4_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); |
| 922 : | suxen_drol | 1.33.2.19 | if (cpu_flags & XVID_CPU_3DNOWEXT) |
| 923 : | suxen_drol | 1.33.2.20 | test_quant(dequant4_inter_c, dequant4_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); |
| 924 : | suxen_drol | 1.33.2.19 | |
| 925 : | emms_mmx(); | ||
| 926 : | |||
| 927 : | #endif | ||
| 928 : | |||
| 929 : | return XVID_ERR_OK; | ||
| 930 : | } | ||
| 931 : | |||
| 932 : | |||
| 933 : | suxen_drol | 1.33.2.13 | int |
| 934 : | xvid_init(void *handle, | ||
| 935 : | int opt, | ||
| 936 : | void *param1, | ||
| 937 : | void *param2) | ||
| 938 : | { | ||
| 939 : | switch(opt) | ||
| 940 : | { | ||
| 941 : | case XVID_INIT_INIT : | ||
| 942 : | return xvid_init_init((XVID_INIT_PARAM*)param1); | ||
| 943 : | |||
| 944 : | case XVID_INIT_CONVERT : | ||
| 945 : | return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); | ||
| 946 : | suxen_drol | 1.33.2.19 | |
| 947 : | case XVID_INIT_TEST : | ||
| 948 : | suxen_drol | 1.33.2.20 | return xvid_init_test((int)param1); |
| 949 : | suxen_drol | 1.33.2.13 | |
| 950 : | default : | ||
| 951 : | return XVID_ERR_FAIL; | ||
| 952 : | } | ||
| 953 : | Isibaar | 1.1 | } |
| 954 : | |||
| 955 : | edgomez | 1.16 | /***************************************************************************** |
| 956 : | * XviD Native decoder entry point | ||
| 957 : | * | ||
| 958 : | * This function is just a wrapper to all the option cases. | ||
| 959 : | * | ||
| 960 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
| 961 : | * else returns the wrapped function result | ||
| 962 : | * | ||
| 963 : | ****************************************************************************/ | ||
| 964 : | |||
| 965 : | edgomez | 1.15 | int |
| 966 : | xvid_decore(void *handle, | ||
| 967 : | int opt, | ||
| 968 : | void *param1, | ||
| 969 : | void *param2) | ||
| 970 : | Isibaar | 1.1 | { |
| 971 : | edgomez | 1.15 | switch (opt) { |
| 972 : | case XVID_DEC_DECODE: | ||
| 973 : | suxen_drol | 1.33.2.11 | return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2); |
| 974 : | edgomez | 1.15 | |
| 975 : | case XVID_DEC_CREATE: | ||
| 976 : | chenm001 | 1.29 | return decoder_create((XVID_DEC_PARAM *) param1); |
| 977 : | edgomez | 1.15 | |
| 978 : | case XVID_DEC_DESTROY: | ||
| 979 : | return decoder_destroy((DECODER *) handle); | ||
| 980 : | Isibaar | 1.1 | |
| 981 : | default: | ||
| 982 : | edgomez | 1.15 | return XVID_ERR_FAIL; |
| 983 : | } | ||
| 984 : | Isibaar | 1.1 | } |
| 985 : | |||
| 986 : | edgomez | 1.16 | |
| 987 : | /***************************************************************************** | ||
| 988 : | * XviD Native encoder entry point | ||
| 989 : | * | ||
| 990 : | * This function is just a wrapper to all the option cases. | ||
| 991 : | * | ||
| 992 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
| 993 : | * else returns the wrapped function result | ||
| 994 : | * | ||
| 995 : | ****************************************************************************/ | ||
| 996 : | Isibaar | 1.1 | |
| 997 : | edgomez | 1.15 | int |
| 998 : | xvid_encore(void *handle, | ||
| 999 : | int opt, | ||
| 1000 : | void *param1, | ||
| 1001 : | void *param2) | ||
| 1002 : | Isibaar | 1.1 | { |
| 1003 : | edgomez | 1.15 | switch (opt) { |
| 1004 : | case XVID_ENC_ENCODE: | ||
| 1005 : | chl | 1.33.2.10 | |
| 1006 : | suxen_drol | 1.20 | if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
| 1007 : | return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, | ||
| 1008 : | (XVID_ENC_STATS *) param2); | ||
| 1009 : | else | ||
| 1010 : | edgomez | 1.15 | return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
| 1011 : | (XVID_ENC_STATS *) param2); | ||
| 1012 : | |||
| 1013 : | case XVID_ENC_CREATE: | ||
| 1014 : | return encoder_create((XVID_ENC_PARAM *) param1); | ||
| 1015 : | |||
| 1016 : | case XVID_ENC_DESTROY: | ||
| 1017 : | return encoder_destroy((Encoder *) handle); | ||
| 1018 : | Isibaar | 1.1 | |
| 1019 : | default: | ||
| 1020 : | edgomez | 1.15 | return XVID_ERR_FAIL; |
| 1021 : | } | ||
| 1022 : | Isibaar | 1.1 | } |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |